Electric power acquisition terminal daily timing error automatic compensation device based on AI technology

Abstract

The invention discloses a daily timing error automatic compensation device of an electric power acquisition terminal based on an AI technology, and relates to the field of electric power automation. The device comprises an error acquisition module (for acquiring parameters such as terminal timing error, environment temperature and power supply voltage, and calculating daily timing error delta T); the system comprises an AI error prediction and compensation model module (based on a random forest, an LSTM and other models, predicting an error trend and generating a compensation strategy), a compensation execution module (a terminal daily timing error adjusting module), a control module (coordinating operation of each module), a data storage module (storing historical data) and a communication module (interacting with a background). Through AI model learning error and multi-factor correlation characteristics, high-precision and full-automatic error compensation is realized, the problems of poor adaptability and low precision of a traditional method are solved, and the method is suitable for AC / DC power acquisition terminals, and is especially suitable for daily timing error calibration in outdoor and multi-environment fluctuation scenes.

Description

Technical Field

[0001] This invention relates to the field of power system automation technology, and in particular to an automatic compensation device for daily timing errors in power acquisition terminals based on artificial intelligence (AI) technology. Background Technology

[0002] Power data acquisition terminals are core equipment for power system metering, monitoring, and control. Their timing accuracy directly impacts the reliability of critical operations such as electricity billing, load dispatching, and fault diagnosis. Currently, power data acquisition terminals primarily rely on internal crystal oscillators for timing. However, crystal oscillators are susceptible to factors such as ambient temperature, voltage fluctuations, and component aging, leading to accumulated daily timing errors (typically ±1 second / day to ±5 seconds / day). Long-term errors can cause metering deviations and data synchronization problems, requiring periodic manual calibration, which is costly and lacks timeliness.

[0003] Chinese patent CN118192188A discloses a clock daily timing error compensation method, device, equipment, and storage medium. The clock daily timing error compensation method includes: obtaining a first temperature value, a second temperature value, a third temperature value, and a fourth temperature value from multiple temperature environments; obtaining daily timing error compensation values ​​for the clock at the first, second, third, and fourth temperature values; performing cubic polynomial curve fitting on the first, second, third, and fourth temperature values ​​and the corresponding daily timing error compensation values ​​to obtain a corresponding first function; and compensating for the clock daily timing error in the real-time temperature environment based on the first function. The drawback of this clock daily timing error compensation method, device, equipment, and storage medium is that it uses fixed parameter correction, making it difficult to adapt to error changes in complex dynamic environments (such as seasonal temperature fluctuations and terminal aging trends). Therefore, there is an urgent need for a technical solution that can adaptively learn error characteristics and achieve high-precision automatic compensation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an automatic compensation device for daily timing errors in power acquisition terminals based on AI technology. By learning the error variation patterns through an AI model, it achieves real-time prediction and dynamic compensation, thereby improving the timing accuracy of power acquisition terminals.

[0005] Technical solution

[0006] An automatic compensation device for daily timing errors in a power data acquisition terminal based on AI technology, comprising:

[0007] 1. Error Acquisition Module: Connected to the timing module of the power acquisition terminal, and simultaneously connected to an external high-precision reference time source (such as a BeiDou / GPS clock or a unified time synchronization signal for the power system), this module acquires the terminal's actual timing data (T1) and reference time data (T0) in real time, and calculates the daily timing error value ΔT = T1 - T0. Optionally, this module can also simultaneously acquire environmental parameters (such as the terminal's internal temperature and power supply voltage) and terminal operating status parameters (such as runtime and CPU load) as auxiliary features for error analysis.

[0008] 2. AI Error Prediction and Compensation Model Module: Communicates with the error acquisition module, receives ΔT and auxiliary feature data, and processes it based on a preset AI model.

[0009] Model training: Construct a training set using historical error data (ΔT sequence), environmental parameters, and runtime parameters, and optimize model parameters (such as the weights of the LSTM network and the decision tree structure of the random forest) through supervised learning;

[0010] Error prediction: Based on the current error value and recent trends, predict the error change curve within a preset period (such as 1 hour or 1 day);

[0011] Compensation strategy generation: Generate compensation amounts (such as crystal oscillator frequency adjustment values ​​and timing register correction values) based on the prediction results to ensure that the timing deviation of the terminal is controlled within the preset threshold (±0.05 seconds / day) after compensation.

[0012] 3. Compensation Execution Module: Connected to both the AI ​​error prediction and compensation model module and the timing module (such as a crystal oscillator circuit or real-time clock chip) of the power acquisition terminal, this module receives the compensation strategy and executes specific adjustments.

[0013] If the terminal supports crystal oscillator frequency calibration, frequency compensation can be achieved by fine-tuning the crystal oscillator control voltage.

[0014] If the terminal only supports register correction, write the compensation value directly to the timing register to correct the accumulated error.

[0015] 4. Control Module: As the core coordination unit of the device, it is connected to the error acquisition module, AI model module and compensation execution module respectively. It configures the working parameters (error acquisition cycle of 1 minute, compensation trigger threshold of 0.05 seconds) and controls the timing logic of each module (such as closed-loop triggering of acquisition → prediction → compensation).

[0016] 5. Data storage module: Stores historical error data, AI model parameters, compensation records, and environmental characteristic data, providing data support for model iteration and fault tracing.

[0017] 6. Communication module: Connects to the power monitoring backend via a power-specific communication protocol (DL / T 645) or wireless communication (4G), uploads error data, compensation results and model status, and receives remote configuration commands from the backend (such as updating model parameters and adjusting compensation thresholds).

[0018] Workflow

[0019] 1. The error acquisition module periodically collects terminal timing data and reference time, calculates ΔT, and synchronously collects environmental parameters;

[0020] 2. The control module determines whether ΔT exceeds the compensation threshold; if it does, the AI ​​model module is triggered.

[0021] 3. The AI ​​model module calls the trained model, inputs the historical ΔT sequence and current environmental parameters, predicts the error trend and generates a compensation strategy;

[0022] 4. The compensation execution module adjusts the terminal timing module according to the strategy and records the adjustment parameters.

[0023] 5. The adjustment results are fed back to the AI ​​model module for online model optimization;

[0024] 6. The data storage module saves all the data throughout the process, and the communication module uploads key information to the backend.

[0025] Beneficial effects

[0026] 1. High-precision compensation: Compared with the traditional fixed parameter method, the AI ​​model can learn the error law under the coupling of multiple factors, and the compensation accuracy is improved to within ±0.1 seconds / day;

[0027] 2. Fully automated operation: From error acquisition to compensation execution, no manual intervention is required throughout the entire process, reducing operation and maintenance costs;

[0028] 3. Adaptive Iteration: Supports online model learning and can adapt to long-term dynamic features such as terminal aging and environmental changes;

[0029] 4. Traceability: Complete records of error data and compensation history facilitate metering audits and fault analysis in the power system. Attached Figure Description

[0030] Figure 1 This is a block diagram of the device structure of the present invention; Figure 2 This is a schematic diagram of the workflow of the present invention. Detailed Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings:

[0032] like Figure 1As shown, the error acquisition module uses an STM32 microcontroller, which reads the real-time clock (RTC) data (T1) from the power acquisition terminal via the UART interface, and simultaneously receives the UTC time (T0) via the Beidou module, calculating ΔT = T1 - T0, with a sampling period of 1 minute. Environmental parameters are acquired through an integrated temperature and humidity sensor (SHT30) and voltage sampling circuit, with resolutions of 0.01℃ and 0.01V, respectively.

[0033] The AI ​​error prediction and compensation model module utilizes an edge computing chip (NVIDIA Jetson Nano) and incorporates an LSTM neural network model. The input layer contains four features (the ΔT sequence over the past 24 hours, current temperature, voltage fluctuation, and the number of days the terminal has been running), the hidden layer has 32 neurons, and the output layer provides the predicted error value and corresponding compensation for the next 24 hours. The model is pre-trained using one year's worth of historical error data (approximately 8760 sets) and trained using the Adam optimizer with mean squared error (MSE) as the loss function.

[0034] The compensation execution module communicates with the terminal's RTC chip (DS3231) via the I2C bus. When the prediction error exceeds 0.5 seconds, it writes a compensation value to the RTC chip (directly correcting the second register). The control module uses an ARM Cortex-M4 processor and coordinates the timing of each module via the SPI bus. It is set to perform a full compensation once a day at 3:00 AM, and trigger temporary compensation when the real-time error exceeds the threshold.

[0035] The data storage module uses 16GB eMMC flash memory, storing data in the format of "time + ΔT + compensation + temperature + voltage," retaining records for the most recent 3 years. The communication module uses a power line carrier chip (HL6300), uploading data to the backend according to the DL / T 645 protocol, with an upload cycle of 1 hour.

[0036] In actual operation, when the terminal's ΔT increases to 0.8 seconds due to a sudden rise in ambient temperature, the control module triggers the AI ​​model module. The model predicts that the error will reach 1.2 seconds in the next 12 hours and generates a strategy of "compensation of +0.1 seconds per hour". The compensation execution module dynamically adjusts the error by modifying the RTC register, and finally controls the cumulative error in 24 hours to within 0.05 seconds.

Claims

1. An automatic compensation device for daily timing errors in a power data acquisition terminal based on AI technology, characterized in that, include: The error acquisition module is used to collect the daily timing error value of the power acquisition terminal in real time; The AI ​​error prediction and compensation model module is connected to the error acquisition module. It is used to receive the daily timing error value, analyze the historical daily timing error data and the current daily timing error value based on the preset AI model, predict the timing error trend of the power acquisition terminal, and generate a corresponding error compensation strategy. The compensation execution module is connected to the AI ​​error prediction and compensation model module and the power acquisition terminal respectively. It is used to adjust the clock chip register parameters of the power acquisition terminal according to the error compensation strategy to realize automatic compensation of daily timing error. The control module is connected to the error acquisition module, the AI ​​error prediction and compensation model module, and the compensation execution module, respectively, and is used to coordinate the runtime sequence and data interaction of each module.

2. The apparatus according to claim 1, characterized in that, The reference time data collected by the error acquisition module comes from the International Standard Time.

3. The apparatus according to claim 1, characterized in that, The AI ​​model in the AI ​​error prediction and compensation model module includes the random forest model, which is trained and optimized using historical daily timing error data and ambient temperature data.

4. The apparatus according to claim 3, characterized in that, The input features of the AI ​​error prediction and compensation model module are matched with the model training data described in claim 3, specifically including: the daily timing error of the power acquisition terminal, real-time ambient temperature, voltage fluctuation, and terminal running time.

5. The apparatus according to claim 1, characterized in that, The compensation execution module achieves error compensation by adjusting the value of the clock chip register inside the power acquisition terminal.

6. The apparatus according to claim 1, characterized in that, The control module is also used to set the error acquisition cycle and the compensation trigger threshold, and store the threshold in the data storage module; when the daily timing error value exceeds the compensation trigger threshold, the AI ​​error prediction and compensation model module is triggered to generate a compensation strategy.

7. The apparatus according to claim 1, characterized in that, It also includes a data storage module, which is connected to the error acquisition module, the AI ​​error prediction and compensation model module and the control module respectively, for storing historical daily timing error data, AI model parameters and compensation records, and responding to data read and write commands from the control module.

8. The apparatus according to claim 1, characterized in that, It also includes a communication module, which is connected to the control module and is used to upload daily timing error data, compensation strategies and compensation results to the power monitoring backend under the coordination of the control module, and to receive configuration instructions from the backend and feed them back to the control module.

9. The apparatus according to claim 1, characterized in that, When adjusting the timing module of the power acquisition terminal, the compensation execution module simultaneously records the daily timing error before and after the adjustment and feeds it back to the AI ​​error prediction and compensation model module (for model iterative optimization) and the data storage module (for historical records).

10. The apparatus according to claim 3, characterized in that, The AI ​​error prediction and compensation model module supports online learning. When a change in the operating environment of the power acquisition terminal is detected, the model parameters are automatically updated to adapt to the new error characteristics.

Images